ChatGPT is becoming the everything app
but it has a problematic personality
I write a newsletter about startups and investing—for ai builders of all levels.
I record mini-tutorials, review tools I’m testing, share my insights and give you a peek behind the digital curtain from an exited founder turned investor.
Hey folks,
Just landed in SF! Writing this in the hotel bar, hopefully it doesnt go wonky 😅
I’m in town meeting founders, investors and friends. Kicking off convos for the builder fund I’ve mentioned - i’ve just signed the docs to open fund II and details coming very soon.
🔎 News worth knowing
ChatGPT’s latest releases for search include making shopping better (items are not ads), message chatgpt in whatsapp and it can search the web, improved citations, trending and auto-complete. ChatGPT feels like its becoming the ‘everything app’ for anyone interacting with the web (vs people via insta/tiktok/x etc). I genuinely rarely use google at all for anything, i can’t see my behaviour reversing.
4o, the default model on ChatGPT, got an update last week, and it started sucking up to absolutely everyone and their dog. OpenAI has temporarily patched it with a system prompt update but I’ve heard the word sycophant enough times for this lifetime. Also, deep research in chatgpt is now available to free users too, but it’s not that deep.
Qwen’s new 30B MoE model has vibes similar to early claude 3.5 sonnet. Qwen3’s family of open weights models was released yesterday with the usual Qwen package of models from 0.6B to 32B parameters. There’s another 235B MoE model with hybrid reasoning but it overthinks and isn’t as good as deepseek r1.
Scale GenAI across your enterprise with proven strategies tested in the real world. Join this exclusive webinar hosted by Fuel iX and featuring Forrester to discover how leading companies are implementing AI copilots, breaking adoption barriers, and achieving frontline success. Get the on-demand recording →*
*sponsored
want to partner with us? Click here
🌐 What I’m consuming
A fully automated end-to-end bug fixing worfklow in cursor (imma have to try this)
and/or this new tool mrge - ai code reviews
An interview with Rahul from Julius AI (i’m an investor).
Ben (not me) made the popular “anatomy of an o1 prompt” image. he has a new workflow for using o3.
One of my latest investments, Smithery, is hiring a founding engineer.
How this guy vibe coded a game (he’s never made one before). I found this v interesting as I’m planning a game for my kids.
Creating a ‘time-travel’ photo app.
Aaron Levie constantly seems switched on with AI - so it was interesting to see what he said about if he was starting a company today; reimagining operations, new business models and ability to do more “long tail” work (ie all the nice to have features)
Building a coding agent from scratch.
How to get the most out of vibe coding.
Duolingo’s CEO goes public with his AI-first plan.
⚙️ Tools I’m tinkering with
Stay competitive in an AI-driven workplace. Gain insights and future-proof your career with the Digital Data Design Institute at Harvard*
Botsheets - turn your google sheets into slides or docs
Lovable 2.0 launched - I haven’t had a chance to play with it yet, reply in the comments how you’re finding it over 1.0
Tiny Agents - a MCP-powered agent in 50 lines of code
I use Flow for voice-to-text on my mac, but my old boss (ryan hoover) has been using Aqua - I’m going to test it, I like the ‘streaming’ feature.
DeepWiki - an encyclopaedia of GitHub repos
Sim Studio - an open-source agent workflow builder
Operative - Let browser agents vibe-test your vibe-coding
Simular - the first AI agent that browses the Internet "with" you (on Mac). Review coming soon.
*sponsored
📈 Gamma’s rise to 50 million users
Written by Shanice
Gamma just hit a huge milestone: 50 million users.
Pretty wild when you think about what they’re building. Gamma is neither a mass-market consumer play, nor viral AI toy, nor enterprise SaaS.
It’s… an AI-powered way to make decks, websites and docs.
Really it’s a productivity tool, aimed at people – knowledge workers, founders, startup teams, educators etc. – who makes slides, share structured ideas, and present things.
It’s horizontal enough to grow big, but not so universal that everyone and their nan is using it to share photos.
It sits in this sort of grey area, which for many tech tools would be their downfall. But it’s a huge part of the reason for Gamma’s success, and makes their new milestone all the more impressive.
Slides don’t suck anymore, yay!
I first came across Gamma around 18 months ago after Ben angel invested (more on that below).
I remember thinking how not awful it felt to use. No fiddly formatting or anaemic-looking slides or decision paralysis about whether to use bullet points or paragraphs of text. Just type a prompt or upload a document, et voilà – a near-perfect, beautiful slide deck or website appears before your very eyes.
It’s not magic, but it certainly feels like it!
Gamma got here with a lean team of 35 and $20M+ in funding.
🍦 Afters
Two london events I’m going to try and attend:
Granola x Vercel hackathon in london: build the silliest productivity app
Cerebral Valley summit (speakers include CEO’s of Uber and Figma)
Lovable’s system prompt is the latest to leak
Dario’s (Anthropic’s cofounder) new essay makes the case for interpretability.
not AI but this is perhaps the coolest car concept i’ve seen ever?
tiny llm - a course of serving LLM on Apple Silicon for systems engineers.
That’s it for today. Feel free to hit reply and share your thoughts. 👋
Enjoy this newsletter? Please forward to a friend.